Photonic Memory Nears Practical Deployment in Integrated Optical Systems

This post contains affiliate links, and I will be compensated if you make a purchase after clicking on my links, at no cost to you.

The dream of fully optical computing has always hit a wall: there just hasn’t been a good, fast photonic memory. Now, researchers at the University of Wisconsin–Madison have built something called the “pLatch.” It’s a photonic memory device that works like optical SRAM, finally plugging a major hole in the optical computing puzzle—and honestly, that’s a big deal for anyone rooting for faster, more scalable data processing.

Closing the Memory Gap in Optical Computing

People have talked about optical computing for decades, promising lightning-fast data because photons move faster than electrons. Optical interconnects and logic have gotten better, but memory stayed electronic, so everything had to be hybrid. That’s always slowed things down.

The new pLatch goes right after that problem. It’s a fully optical memory element that fits into today’s silicon photonics. No more shuttling data between light and electrons just for temporary storage.

From Concept to Practical Photonic SRAM

The pLatch acts like an optical version of static random-access memory (SRAM), the high-speed memory found in processor caches. Regular electrical SRAM runs at about 2–3 GHz. The Wisconsin team’s photonic version? It’s hitting:

  • Write speeds close to 20 GHz
  • Read speeds up to 50–60 GHz
  • Those numbers blow past traditional SRAM. Photons just have an edge in high-bandwidth stuff.

    How the pLatch Works: Standard Silicon Photonics

    The pLatch isn’t just fast—it’s actually practical. Instead of weird materials or experimental processes, it uses the same parts you’ll find in silicon photonics today.

    Here’s what’s inside:

  • Photodiodes that convert and control light signals
  • Micro-ring resonators to trap and tweak certain wavelengths of light
  • Optical waveguides that route light around the chip
  • Compatibility with Existing Semiconductor Foundries

    The pLatch is built for standard silicon photonics process flows, so it works with big commercial foundries like GlobalFoundries. That’s a huge step if this is going to move from the lab to the real world.

    The project team worked closely with AIM Photonics and GlobalFoundries. They made sure layouts, device shapes, and materials could all be made using industry-standard tools and rules. This isn’t just a lab curiosity—it’s ready for manufacturing.

    Why Photonic Memory Matters for High-Performance Systems

    Data centers and high-performance computing (HPC) setups already use optical links to move data between racks. But inside chips and modules, we’re still stuck with electronics for memory and logic, which creates annoying conversion bottlenecks.

    The pLatch changes that by making possible:

  • Fully optical data paths from interconnects right into memory
  • Lower latency since you don’t have to keep converting between light and electricity
  • More bandwidth for workloads that hammer memory
  • Ideal for Data Centers and Large-Scale Connectivity

    Photonic parts are still bigger than electronic ones, so you won’t see these in your phone anytime soon. But in big systems where space isn’t so tight, they make a lot of sense.

    Here, the pLatch can act as a super-fast buffer or cache between optical interconnects and processors. That’s a win for:

  • High-performance computing (HPC)
  • AI and machine learning clusters
  • Cloud data center fabrics and disaggregated memory systems
  • From Lab to Conference Stage: Validating the Breakthrough

    Akhilesh Jaiswal and PhD student Md Abdullah-Al Kaiser led the work, along with their collaborators. They didn’t just prove the physics—they set out to show this could be made at scale.

    The team presented at the International Electron Devices Meeting (IEDM) 2025, a top spot for big advances in semiconductor devices. That’s a pretty strong signal the tech is ready for serious attention from both academia and industry.

    A Pivotal Step Toward Practical Optical Computing

    The pLatch offers a high-speed, fully optical memory solution. It uses standard silicon photonics processes, which is pretty significant if you care about practical applications.

    This device tackles one of the toughest challenges in optical computing. Sure, integrating big arrays and pairing them with optical logic and interconnects will take more work, but we’ve got a real starting point now.

    If researchers can scale this technology, it might transform how we build data centers and HPC systems. Imagine architectures that run on photonics at their core, pushing speeds well past what electronics can manage alone.

     
    Here is the source article for this story: Photonic memory moves closer to practical deployment

    Scroll to Top